Firecrawl CLI - CLI Agent Audit
Generated: 2026-05-27
CLI version: 1.18.1
Scope: Critical failure modes
Summary
Firecrawl CLI is installable and has broad human-facing docs, but its current agent contract is weak on structured errors, non-interactive auth, safe side effects, and machine-readable self-description.
| Score | Result |
|---|---|
| Failure mode score | 0.48/3 (0 pass, 9 partial, 12 fail, 2 indeterminate) |
| Readiness score | 7/15 [C] |
Key Findings
- Auth-required commands can exit 0 while printing an interactive login prompt to stdout.
--jsondoes not reliably produce JSON on auth and error paths.- Browser login prints a URL but then waits indefinitely for auth in headless execution without a structured fallback.
- There is no schema/manifest for commands, exit codes, scopes, safe defaults, or interactivity requirements.
- Side-effecting setup/config commands lack dry-run/effect metadata.
Files
| File | Purpose |
|---|---|
| report-index.md | Report index and score summary |
| report-issues.md | Concrete bugs and failure-mode gaps |
| report-runtime.md | Operational brief for agents invoking the CLI |
| report-agent-dev.md | Integration guide for agent builders |
| report-dev.md | Fix list for CLI authors |
| findings.md | Per-failure-mode scores |
| trace.md | Raw check trace summaries |
| readiness.md | Proactive readiness score |
Generated by cli-agent-audit - CLI Agent Spec